TensorFlow
TensorFlow was developed by Google. The initial version of it was released under the Apache Licence 2.0.
It is an open source machine learning software library developed by Google to solve the problem of building and training a neural network to automatically find and classify images, achieving human perceptual quality. TensorFlow is a solid solution for complex projects such as building multilayer neural networks. It is used for voice or image recognition and word processing applications such as Google Translate.
Clients: Airbnb, Coca-Cola, DeepMind, Google, Intel, Twitter, Nersc, GE Healthcare и т.д.
Use cases
- Image analysis to fight fake photos (Airbnb)
Solution: Creation of an algorithm based on computer vision and deep learning technology that can be used in solving problems with fake photos in listings on the website.
Result: With these algorithm-detected amenities, we are able to verify the quality of the listings from hosts and make it much easier for guests to find homes with specific amenity needs. This will push the frontier of photo intelligence at Airbnb to the next level.
- Neural network for detecting anomalies on MRI (GE Healthcare)
Solution: Application of TensorFlow to train a neural network to identify specific anatomy during a brain magnetic resonance imaging (MRI) exam. Development of a framework for intelligent MRI slice placement (ISP).
Results: Tests have shown that AI helped to reduce the time required for the scan operator to locate anomalies on an MRI-scan by 40%-60%. Moreover, a decrease in errors and improved accuracy has been seen, which can lead to fewer repeat patient examinations and higher diagnostic quality.
Benefits
TensorFlow is an excellent machine learning framework for when large amounts of data need to be processed and/or when large deep neural networks need to be trained.
Building neural networks on TensorFlow becomes easier with the Keras API.
